Gemini 3.8 Live Extended Thinking
Google · Released Sep 2026
Gemini 3.8 Live Extended Thinking is the voice-agent model to start with: it tops Artificial Analysis' Speech to Speech Index and reasons and calls tools in the background while it keeps talking, at the same rate as the default Gemini 3.8 Live. The caveat is the Live API surface: no structured outputs, caching, code execution or batch. Right for teams building real-time voice agents that must solve multi-step problems mid-call.
Is it right for you?
Good for
- Real-time voice agents that must reason through multi-step problems mid-conversation
- Calling tools and APIs in the background while the conversation continues
- Multilingual voice: Google says it detects and switches between 97 languages mid-conversation
Not good for
- Workloads that need structured outputs, context caching, code execution or the Batch API, none of which the model supports
- Inputs above 131,072 tokens
- Sensitive data on the free tier, which Google's pricing page marks as used to improve its products
How it performs by task
Real-time voice agents
Top Artificial Analysis Speech to Speech Index score (82.6) with background reasoning
Agentic tool use during calls
Async function calling while audio streams; vendor-reported 68.6% on τ-Voice
Structured data extraction
Structured outputs are not supported on this model
Pricing
Input
$0.75 / 1M (text)
Output
$4.50 / 1M (text)
Context
131K in, 65K out; audio $3/$12 per 1M
Benchmarks
No verdict changes yet
The clock starts day one. Changes land here as our verdict evolves.
Sources
Verification log
- Pricing· No changes
Automated agent
- Pricing· No changes
Automated agent